The Static Magnetic Field Dependence of Chemical Exchange

can depend on the proper choice of positive or negative square roots in eqs ... Rb and Δω ) 628 s-1 (100 Hz), but the shapes of the resulting curves...
5 downloads 8 Views 255KB Size
Download PDF
J. Am. Chem. Soc. 2000, 122, 2867-2877

2867

The Static Magnetic Field Dependence of Chemical Exchange Linebroadening Defines the NMR Chemical Shift Time Scale Oscar Millet,† J. Patrick Loria,‡ Christopher D. Kroenke,‡ Miquel Pons,† and Arthur G. Palmer, III*,‡ Contribution from the Departament de Quı´mica Orga` nica, UniVersitat de Barcelona, Martı´ i Franque` s 1-11, E-08028 Barcelona, Spain, and Department of Biochemistry and Molecular Biophysics, Columbia UniVersity, 630 West 168th Street, New York, New York 10032 ReceiVed September 29, 1999. ReVised Manuscript ReceiVed January 7, 2000

Abstract: The static magnetic field dependence of chemical exchange linebroadening in NMR spectroscopy is investigated theoretically and experimentally. Two-site exchange (A / B) is considered with site A more highly populated than site B (pa > pb), a shift difference between sites equal to ∆ω, and an exchange rate constant given by kex. The exchange contribution to the transverse relaxation rate constant for the more highly populated site is denoted Rex. The dependence of Rex on the static magnetic field strength is characterized by a scaling parameter R ) d ln Rex/d ln ∆ω, in which 0 e R e 2 for pa > 0.7. The value of R depends on the NMR chemical shift time scale for the exchange process: for slow exchange (kex/∆ω < 1), 0 e R < 1; for intermediate exchange (kex/∆ω ) 1), R ) 1; and for fast exchange (kex/∆ω > 1), 1 < R e 2. Consequently, the static magnetic field dependence of Rex defines the chemical shift time scale for an exchange process even if the populations are so highly skewed (pa . pb) that the minor resonance is not observable in the slow exchange limit. The theoretical results are verified by measuring 15N transverse relaxation rate constants at static magnetic fields of 11.7 and 14.1 T and temperatures of 300 and 313 K for the protein basic pancreatic trypsin inhibitor. At each combination of static magnetic field and temperature, the rate constants were measured using Carr-Purcell-Meiboom-Gill and Hahn echo techniques with spin-echo delays ranging from 1.0 to 64.5 ms. 15N resonances for residues in the region of the Cys14-Cys38 disulfide bond are broadened due to chemical exchange. Values of R obtained from the relaxation rate constants range from 0.26 ( 0.17 for Arg39 at 300 K to 1.96 ( 0.25 for Cys38 at 313 K. For Cys38 and Arg39, the two residues most strongly affected by chemical exchange, values of kex were determined to be 380 ( 70 s-1 and 530 ( 90 s-1 at 300 K and 1300 ( 290 s-1 and 1370 ( 160 s-1 at 313 K by global analysis of the relaxation rate constants. The scaling parameters R indicate that chemical exchange for most residues in basic pancreatic trypsin inhibitor does not satisfy kex/∆ω . 1. Consequently, the assumption of fast-limit quadratic scaling of exchange broadening in proteins and other macromolecules may be incorrect, even if a single broadened resonance is observed for a nuclear spin. The theoretical results for the static magnetic field dependence of chemical exchange broadening in NMR spectroscopy are applicable to other nuclei and to other techniques for measuring chemical exchange linebroadening.

Introduction Knowledge of molecular dynamics is essential for understanding the biophysical properties and biological functions of proteins.1,2 NMR spin relaxation measurements have proven to be a powerful tool for the characterization of dynamic processes in proteins in solution over a wide range of time scales.3,4 Fast motions on picosecond-nanosecond time scales that modulate the chemical shift, dipolar coupling, and quadrupolar coupling can be characterized by heteronuclear (2H, 13C, and 15N) spin relaxation NMR spectroscopy using established experimental protocols.5 Recent applications of these techniques have focused * Author to whom correspondence should be addressed. Telephone: (212) 305-8675. Fax: (212) 305-7932). E-mail: [email protected]. † Universitat de Barcelona. ‡ Columbia University. (1) Karplus, M.; McCammon, J. A. Annu. ReV. Biochem. 1983, 53, 263300. (2) Frauenfelder, H.; Sligar, S. G.; Wolynes, P. G. Science 1991, 254, 1598-1603. (3) Palmer, A. G. Curr. Opin. Struct. Biol. 1997, 7, 732-737. (4) Kay, L. E. Nat. Struct. Biol. 1998, 5, 513-517.

on the role of conformational entropy in ligand binding6,7 and protein folding.8,9 Chemical or conformational kinetic processes on microsecond-millisecond time scales that stochastically transfer nuclear spins between magnetic environments with different isotropic chemical shifts, referred to generically as chemical exchange, also can be studied by NMR spectroscopy.5 Chemical exchange contributes to the transverse relaxation rate in the laboratory frame (R2) and in the rotating frame (R1F). Consequently, chemical exchange can be characterized by measuring the excess contribution to the transverse relaxation rate constant, commonly called Rex. Although experimental methods for characterizing motions on microsecond-milli(5) Palmer, A. G.; Williams, J.; McDermott, A. J. Phys. Chem. 1996, 100, 13293-13310. (6) Kay, L. E.; Muhandiram, D. R.; Wolf, G.; Shoelson, S. E.; FormanKay, J. D. Nat. Struct. Biol. 1998, 5, 156-163. (7) Bracken, C.; Carr, P. A.; Cavanagh, J.; Palmer, A. G. J. Mol. Biol. 1999, 285, 2133-2146. (8) Yang, D.; Kay, L. E. J. Mol. Biol. 1996, 263, 369-382. (9) Alexandrescu, A. T.; Rathgeb-Szabo, K.; Rumpel, K.; Jahnke, W.; Schulthess, T.; Kammerer, R. A. Protein Sci. 1998, 7, 389-402.

10.1021/ja993511y CCC: $19.00 © 2000 American Chemical Society Published on Web 03/14/2000

2868 J. Am. Chem. Soc., Vol. 122, No. 12, 2000

Millet et al.

second time scales are not as well-established as the experimental methods for faster motions, a number of new techniques have been developed that are reinvigorating the investigation of chemical exchange in proteins.10-15 Recent applications of these techniques include investigations of ligand binding,16 loop and domain motions in enzymatic catalysis,17 and protein conformational changes.14,18 For simplicity in the following, only the two-site exchange reaction is considered: k1

A {\ }B k -1

(1)

in which the exchange rate constant, kex, is defined as19

kex ) k1 + k-1 ) k1/pb ) k-1/pa

(2)

pa is the equilibrium population of site A, pb is the equilibrium population of site B, pa + pb ) 1, k1 is the forward first-order kinetic rate constant, and k-1 is the reverse first-order kinetic rate constant. The two sites are assumed to have distinct chemical shifts ωa and ωb, respectively. The frequency difference between the chemical shifts of the two sites is ∆ω ) |ωa - ωb| ) |γ∆σB0|, in which γ is the gyromagnetic ratio for the exchanging nuclear spin, ∆σ is the difference in chemical shielding of the two sites, and B0 is the static magnetic field strength. The transverse relaxation rate constants for sites A and B in the absence of conformational exchange are denoted Ra and Rb, respectively. The contribution to the line width or transverse relaxation rate constant from a chemical kinetic process depends critically on whether the exchange process is slow (kex/∆ω < 1), intermediate (kex/∆ω ≈ 1), or fast (kex/∆ω > 1) on the NMR chemical shift time scale.20 If the populations of the two sites are similar, then slow exchange is recognized easily by the presence of two resolved resonances with frequencies ωa and ωb, while fast exchange is recognized by the presence of a single averaged resonance with frequency paωa + pbωb. Unfortunately, as emphasized by Ishima and Torchia,12 in many cases of interest, the populations of the sites are highly unequal. For example, if site A is more stable than site B by only 2kBT, in which kB is the Boltzmann constant, then pa ) 0.88 and pb ) 0.12. In the slow exchange limit, the resonance at ωb is both lower intensity, by a factor pb/pa, and significantly broader, by a factor (Rb + pakex)/(Ra + pbkex), than the resonance at ωa. As a result, if pa . pb, then the resonance at ωb may be undetectable. Thus, the mere observation of a single exchangebroadened resonance does not necessarily indicate that the exchange process is fast on the chemical shift time scale. (10) Akke, M.; Palmer, A. G. J. Am. Chem. Soc. 1996, 118, 911-912. (11) Zinn-Justin, S.; Berthault, P.; Guenneugues, M.; Desvaux, H. J. Biomol. NMR 1997, 10, 363-372. (12) Ishima, R.; Wingfield, P. T.; Stahl, S. J.; Kaufman, J. D.; Torchia, D. A. J. Am. Chem. Soc. 1998, 120, 10534-10542. (13) Konrat, R.; Tollinger, M. J. Magn. Reson. 1999, 13, 213-221. (14) Loria, J. P.; Rance, M.; Palmer, A. G. J. Am. Chem. Soc. 1999, 121, 2331-2332. (15) Mulder, F. A. A.; van Tilborg, P. J. A.; Kaptein, R.; Boelens, R. J. Biomol. NMR 1999, 13, 275-288. (16) Evena¨s, J.; Forse´n, S.; Malmendal, A.; Akke, M. J. Mol. Biol. 1999, 289, 603-607. (17) Ishima, R.; Freedberg, D.; Wang, Y.-X.; Louis, J. M.; Torchia, D. A. Structure 1999, 7, 1047-1055. (18) Akke, M.; Liu, J.; Cavanagh, J.; Erickson, H. P.; Palmer, A. G. Nat. Struct. Biol. 1998, 5, 55-59. (19) Davis, D. G.; Perlman, M. E.; London, R. E. J. Magn. Reson., Ser. B 1994, 104, 266-275. (20) Cavanagh, J.; Fairbrother, W. J.; Palmer, A. G.; Skelton, N. J. Protein NMR Spectroscopy: Principles and Practice; Academic Press: San Diego, 1996; p 587.

In the following, the dependence of the exchange contribution to the transverse relaxation rate constant on the static magnetic field strength is investigated. In contrast to statements in the literature that exchange broadening depends quadratically on the static magnetic field strength,21-24 Rex for the more highly populated site A is demonstrated to depend on the static magnetic field strength through the scaling relationship δRex/ Rex ) R δB0/B0, in which 0 e R e 2 for pa > 0.7. The value of R depends on the NMR chemical shift time scale for the exchange process. Quadratic scaling of Rex is obtained only for R f 2, and Rex is independent of the static magnetic field for R f 0. Thus, the static magnetic field dependence of Rex defines the chemical shift time scale for a chemical exchange process even if the populations are highly skewed with pa . pb. The theoretical results are verified using 15N Carr-PurcellMeiboom-Gill (CPMG)25,26 measurements of transverse relaxation for basic pancreatic trypsin inhibitor (BPTI). A conformational exchange process in the region of the Cys14Cys38 disulfide bond has been studied extensively by NMR spectroscopy.14,27-29 Over the temperature range from 300 to 313 K, most of the exchange-broadened resonances in BPTI have values of R between 0.5 and 1.6; thus, exchange is neither very slow (kex/∆ω , 1) nor very fast (kex/∆ω . 1) on the chemical shift time scale. The results suggest that the assumption of fast-limit quadratic scaling of exchange broadening in proteins and other macromolecules generally is not tenable. Theory The contributions to transverse relaxation rate constants from chemical exchange processes in macromolecules most commonly are measured using CPMG14,30 or R1F measurements.10,11,28 The theoretical treatment of chemical exchange is more highly developed for the CPMG experiment19,31 than for the R1F experiment.32,33 In addition, the experimental results reported herein use the CPMG method to measure the effects of chemical exchange on 15N transverse relaxation rate constants in BPTI. Accordingly, the theoretical presentation below focuses on the CPMG technique; however, similar conclusions regarding the static magnetic field dependence of conformational exchange are applicable to R1F measurements. A general expression for the phenomenological transverse relaxation rate constant for site A, R2(1/τcp), is given by,19,31

R2(1/τcp) ) 1

(

/2 Ra + Rb + kex -

)

1 cosh-1[D+ cosh(η+) - D- cos(η-)] τcp (3)

in which τcp is the delay between 180° pulses in the CPMG (21) Farrow, N. A.; Zhang, O.; Szabo, A.; Torchia, D. A.; Kay, L. E. J. Biomol. NMR 1995, 6, 153-162. (22) Mandel, A. M.; Akke, M.; Palmer, A. G. J. Mol. Biol. 1995, 246, 144-163. (23) Peng, J.; Wagner, G. Biochemistry 1995, 34, 16733-16752. (24) Phan, I. Q. H.; Boyd, J.; Campbell, I. D. J. Biomol. NMR 1996, 8, 369-378. (25) Carr, H. Y.; Purcell, E. M. Phys. ReV. 1954, 94, 630-638. (26) Meiboom, S.; Gill, D. ReV. Sci. Instrum. 1958, 29, 688-691. (27) Otting, G.; Liepinsh, E.; Wu¨thrich, K. Biochemistry 1993, 32, 35713582. (28) Szyperski, T.; Luginbu¨hl, P.; Otting, G.; Gu¨ntert, P.; Wu¨thrich, K. J. Biomol. NMR 1993, 3, 151-164. (29) Beeser, S. A.; Goldenberg, D. P.; Oas, T. G. J. Mol. Biol. 1997, 269, 154-164. (30) Orekhov, V. Y.; Pervushin, K. V.; Arseniev, A. S. Eur. J. Biochem. 1994, 219, 887-896.

Definition of NMR Chemical Shift Time Scale

J. Am. Chem. Soc., Vol. 122, No. 12, 2000 2869 slow and fast exchange are

Rex ) pbkex

(kex/∆ω f 0)

(9)

Rex ) papb∆ω2/kex

(kex/∆ω f ∞)

(10)

respectively. The limiting results that Rex depends quadratically on the static field strength for very fast exchange21,23,24 and that Rex is independent of the static field strength for very slow exchange34 are known. The functional form of Rex given by eq 8 is not overly sensitive to the values of Ra and Rb provided that |Ra - Rb|/∆ω , (pa - pb). In the case of interest herein, pa . pb and assuming Ra ) Rb:

Rex ≈

Figure 1. Chemical exchange and the NMR chemical shift time scale. The values of (a) Rex, calculated from eq 8, and (b) R, calculated from eq 13, are shown as functions of kex. Values of Rex are normalized by the value of Rex at coalescence (kex/∆ω ) 1), and the x-axis is normalized by ∆ω for presentation. The slow exchange regime corresponds to kex/∆ω 1. Curves are drawn for site populations pa f 1 (heavy line), 0.9, 0.8, and 0.7. Calculations were performed assuming Ra ) Rb and ∆ω ) 628 s-1 (100 Hz), but the shapes of the resulting curves depend weakly on these assumptions. The results for R with pa f 1 are identical to eq 14.

[

]

ψ + 2∆ω2 D( ) /2 ( 1 + 2 (ψ + ζ2)1/2 η( )

τcp

(4)

[(ψ + (ψ2 + ζ2)1/2]1/2

(5)

x2

ψ ) (Ra - Rb - pakex + pbkex)2 - ∆ω2 + 4pa pbk2ex (6) ζ ) 2∆ω(Ra - Rb - pa kex + pb kex)

(7)

Evaluation of these equations for particular sets of parameters can depend on the proper choice of positive or negative square roots in eqs 4 and 5. The variation of R2(1/τcp) as a function of 1/τcp is called relaxation dispersion. The chemical exchange contribution to transverse relaxation, Rex, is defined as the difference between the apparent relaxation rate constants in the slow and fast pulsing limits:19

Rex ) ∆R2(0,∞) ) R2(1/τcp f 0) - R2(1/τcp f ∞) ) 1 1 /2 (ψ + ∆ω2)1/2 [ψ + (ψ2 + ζ2)1/2]1/2 x2

{

}

(8)

in which ∆R2(1/τcp1,1/τcp2) is the difference between the transverse relaxation rate constants for two values of 1/τcp. Figure 1a presents calculated values of Rex as a function of kex using eq 8, assuming Ra ) Rb. The limiting values of Rex in (31) Jen, J. J. Magn. Reson. 1978, 30, 111-128. (32) Meiboom, S. J. Chem. Phys. 1961, 34, 375-388. (33) Deverell, C.; Morgan, R. E.; Strange, J. H. Mol. Phys. 1970, 18, 553-559.

1 + (kex/∆ω)2

(11)

For small changes in the static magnetic field, the fractional change in the chemical exchange broadening, δRex/Rex, and the fractional change in the static field, δB0/B0, are related by

δB0 δRex )R Rex B0

(12)

The constant of proportionality or scaling factor R is defined by

R)

pulse train, 1

pa pbkex

d ln Rex d ln ∆ω

(13)

Provided that pa . pb, R satisfies the constraints 0 e R e 2, which generalizes the limiting cases of eqs 9-10. For example, using eqs 11 and 13,

R)

2(kex/∆ω)2 1 + (kex/∆ω)2

(14)

in the limit pa f 1. As shown by eq 14, R is a function only of kex/∆ω and therefore defines the NMR chemical shift time scale:

0eR 1. Calculations were performed for values of ∆ω equal to (a) 157 s-1 (25 Hz), (b) 314 s-1 (50 Hz), and (c) 628 s-1 (100 Hz). All calculations assumed Ra ) Rb.

of τcp. In the present work, values of R2(1/τcp) for τcp ) 64.5 ms and τcp ) 1.0 ms were used to define

R′ ) d ln ∆R2(0.016 ms-1, 1.00 ms-1)/d ln ∆ω (16) As shown in Figure 2, R′ is a good approximation to R only if pa . pb and ∆ω is sufficiently large. However, a value of R′ < 1 is never observed if kex/∆ω > 1. Therefore, a value of R′ < 1 always indicates that the system is in slow exchange, but a value of R′ > 1 does not definitively indicate the chemical shift time scale. The complexity of eq 3 obfuscates the form of the dependence of R2(1/τcp) on 1/τcp. Ishima and Torchia have proposed a simple function that approximates eq 3 over all time scales provided pa . pb:34

R2(1/τcp) ) R2(1/τcp f ∞) + papb∆ω2kex/[k2ex + (p2a ∆ω4 + 144/τ4cp)1/2] (17) A simple measure of the τcp-dependence of relaxation dispersion is given by the value of 1/τcp for which R2(1/τcp) ) [R2(1/τcp f

The 15N R2(1/τcp) measurements were performed at two static magnetic fields, B0 ) 11.7 T (corresponding to a proton Larmor frequency of 500.13 MHz) and 14.1 T (corresponding to a proton Larmor frequency of 600.13 MHz), and two temperatures, T ) 300 K and 313 K, using a 2.6 mM BPTI (U-98% 15N) sample in 90% H2O/ 10% D2O. Data were recorded at 11.7 T using a Bruker DRX-500 spectrometer and at 14.1 T using a Bruker DRX-600 spectrometer. Both spectrometers were equipped with triple-resonance three-axis gradient probeheads. Sample temperatures were calibrated using an 80% ethylene glycol sample in d6-DMSO and a calibration curve provided by Bruker Instruments. The relaxation-compensated CPMG pulse sequence14 shown in Figure 3a was used to record CPMG experiments for most values of τcp. The conventional pulse sequence35 for measuring 15N R2 shown in Figure 3b was used to record the CPMG experiments for values of τcp ) 10.8 ms and 21.5 ms. Both sequences average in-phase and antiphase coherences in order to eliminate any τcp-dependent effects arising from differential relaxation of these coherences and apply 1H 180° pulses to eliminate 1H-15N dipolar/15N chemical shift anisotropy relaxation interference.36,37 In contrast to the sequence of Figure 3a, the sequence of Figure 3b begins with a refocused INEPT element to generate inphase magnetization at the beginning of the CPMG period so that the 1H 180° pulses used to eliminate relaxation interference are applied only when the 15N coherence is in-phase. In the sequence of Figure 3a, the averaging of coherences is performed explicitly during the U period; in the sequence of Figure 3b, the averaging is performed implicitly by setting τcp ) m/JNH, in which JNH ) 93 Hz is the average value of the one-bond 1H-15N scalar coupling constant and m ) 1 or 2.37 The sequence of Figure 3b is advantageous for values of τcp ) 10.8 ms and 21.5 ms because the first nonzero time point can be recorded at 2τcp rather than 4τcp as necessary for the sequence of Figure 3a. At each combination of τcp, T and B0, R2(1/τcp) was determined from a time series in which 4nτcp for the sequence of Figure 3a or 2nτcp for the sequence of Figure 3b were varied between 0 and 320 ms by varying the integer n. Typically, 9-10 individual time points were recorded plus two duplicate time points. The time points for n ) 0 were recorded by omitting the bracketed segments in Figure 3a and 3b. The value of R2(1/τcp) for τcp ) 64.5 ms was obtained using the Hahn-echo pulse sequence shown in Figure 3c. Only two time points were recorded to determine the relaxation rate constant: one at zero time obtained with the bracketed segments omitted, and one at τcp ) 64.5 ms. The value of τcp was set to 6/JNH to ensure that the 1H 180° (35) Farrow, N. A.; Zhang, O.; Forman-Kay, J. D.; Kay, L. E. Biochemistry 1995, 34, 868-878. (36) Kay, L. E.; Nicholson, L. K.; Delaglio, F.; Bax, A.; Torchia, D. A. J. Magn. Reson. 1992, 97, 359-375. (37) Palmer, A. G.; Skelton, N. J.; Chazin, W. J.; Wright, P. E.; Rance, M. Mol. Phys. 1992, 75, 699-711.

Definition of NMR Chemical Shift Time Scale

Figure 3. Pulse sequences for measuring 15N R2(1/τcp). (a) Relaxation compensated CPMG experiment14 for recording data with τcp < 10 ms. (b) CPMG experiment for recording data with τcp ) 10.8 or 21.6 ms.35 (c) Hahn-echo pulse sequence for recording data with τcp ) 64.5 ms. In (a), averaging of the relaxation rate constants for in-phase and antiphase coherences is performed by interchanging the coherences during U.14 In (b), averaging is performed by setting τcp ) m/JNH, in which m is an integer and JNH is the one-bond 1H-15N scalar coupling constant;37 m ) 1 for τcp ) 10.8 ms and m ) 2 for τcp ) 21.6 ms. In (c), averaging is performed by setting τcp ) 2m/JNH; m ) 3 for τcp ) 64.5 ms. Relaxation delays are (a) t ) 4nτcp and (b) t ) 2nτcp, in which n g 0 is an integer; decay curves are obtained by recording a series of spectra for different values of n. In (c) spectra are recorded only at values of t ) 0 and t ) τcp. Spectra for t ) 0 are obtained by omitting the bracketed elements in (a-c). Narrow and wide bars depict 90° and 180° pulses, respectively. All pulses are x phase unless otherwise indicated. 15N decoupling during acquisition was achieved with a 2 kHz GARP sequence.51 Delays are τ ) τcp/2, ∆ ) 2.7 ms, δ1 ) 750 µs, δ2 ) 2.05 ms, δ3 ) 384 µs. The phase cycle is φ1 ) x, -x; φ2 ) 4(x), 4(-x); φ3 ) x, x, y, y, -x, -x, -y, -y; φ4 ) x, -x; receiver ) x, -x, -x, x. The gradients are sine shaped with duration G1-G7 ) 1, 0.4, 2, 0.5, 1.8, 0.6, and 0.184 ms and amplitude G1xyz ) 8; G2xyz ) 6; G3z ) 15; G4xyz ) 7; G5xyz ) 24; G6z ) 21; and G7xyz ) 24 G/cm. Coherence selection was achieved by inverting the amplitude of G5 and phase φ4;52 φ1 and the receiver phases were inverted for each t1 increment to shift axial peaks to the edge of the spectrum.53 pulses used to eliminate 1H-15N dipolar/15N chemical shift anisotropy relaxation interference are applied when the 15N coherences are inphase. The zero time point must be acquired with the 1H 180° pulses removed from the sequence; otherwise, these pulses would be applied to antiphase coherence, and imperfections in the pulses would lead to loss of signal intensity compared with spectra acquired with τcp ) 64.5 ms. At a temperature of 300 K and B0 ) 11.7 T, data were recorded for τcp values of 1.0, 1.5, 2.0, 4.0, 6.6, 10.8, 21.5, and 64.5 ms; the data at

J. Am. Chem. Soc., Vol. 122, No. 12, 2000 2871 τcp ) 1.0, 1.5, and 2.0 ms were reported earlier and used a variant of the pulse sequence in Figure 3a.14 At each other combination of temperature and static field strength, data were collected for τcp values of 1.0, 1.5, 2.0, 4.0, 6.6, 10.8, and 65.4 ms. Spectra were acquired using (128 × 2048) complex points and spectral widths of (2500 × 12500) Hz in the (t1 × t2) dimensions. The 1H carrier frequency was set coincident with the water resonance; the 15N carrier frequency was set to 115.5 ppm. The recycle delay was 3 s. A total of 16 transients were recorded for each complex point for spectra recorded using the pulse sequences of Figure 3a and 3b. For the pulse sequence of Figure 3c, the spectrum at time zero was recorded with a total of 16 transients for each complex point while the spectrum at τcp ) 65.4 ms was recorded as 8 duplicate spectra with 16 transients per complex point. FELIX97 (MSI) was used for the processing of all spectra. The free induction decays were processed in F2 by applying a convolution filter to suppress the solvent signal, apodizing with an exponential window function, zero-filling once, and Fourier transforming. The resulting spectra were phase-corrected and baseline-corrected. The F1 interferograms were apodized with a Kaiser window function, zero-filled once, Fourier-transformed, and phase-corrected. Peak intensities were measured as the sum of the intensities for a 3 × 3 grid centered on the peak maxima. Time series data obtained using Figure 3a and 3b were fit using the in-house program CurveFit to a monoexponential decay function, I(t) ) I(0) exp[-R2(1/τcp)t], in which I(t) is the intensity of the resonance in the spectrum recorded at time t ) 4nτcp or t ) 2nτcp for data recorded with the sequence of Figure 3a or 3b, respectively. Experimental uncertainties in the peak intensities were estimated from duplicate spectra.38,39 Uncertainties in the fitted rate constants were obtained from jackknife simulations.40 For the Hahn-echo experiment of Figure 3c, rate constants were obtained from R2(1/τcp) ) ln[I(0)/〈I(τcp)〉]/τcp, in which τcp ) 64.5 ms, I(0) is the intensity in the spectrum recorded with a relaxation delay of zero, and 〈I(τcp)〉 is the trimmed mean intensity calculated after eliminating the maximum and minimum intensities from the 8 spectra recorded at the nonzero relaxation delay. Uncertainties in 〈I(τcp)〉 were taken to be the standard error in the trimmed mean. I(0) was assumed to have the same relative uncertainty as the trimmed set of I(τcp) values. The experimental uncertainties in R2(1/τcp) determined by this method were confirmed by recording duplicate data sets at T ) 313 K and B0 ) 11.7 T and at T ) 300 K and B0 ) 14.1 T. Dispersion curves for R2(1/τcp) were fit to eq 3 or to the fast-limit formula:41

R2(1/τcp) ) R2(1/τcp f ∞) + Rex[1 - 2 tanh(kexτcp/2)/(kexτcp)] (19) All data points were weighted equally in performing the curve-fitting, using the mean uncertainty for the R2(1/τcp) values in the dispersion curve and uncertainties in fitted parameters were obtained from jackknife simulations. The value of R2(1/τcp f ∞) obtained from the pulse sequences in Figure 3 is the average of the relaxation rate constants for in-phase and antiphase 15N magnetization.14 Values of R were calculated from values of Rex measured for two values of B0 using the following expression:

R)

(

)(

)

B02 + B01 Rex2 - Rex1 B02 - B01 Rex2 + Rex1

(20)

in which Rex1 and Rex2 are the values measured at the lower field, B01 (11.7 T) and the higher field, B02 (14.1 T). The value of R calculated by eq 20 is a numerical approximation to eq 13 and is associated with an effective field (B01 + B02)/2. Values of R′ were obtained by substituting ∆R2(0.016 ms-1, 1.00 ms-1) for Rex in eq 20. Uncertainties (38) Palmer, A. G.; Rance, M.; Wright, P. E. J. Am. Chem. Soc. 1991, 113, 4371-4380. (39) Skelton, N. J.; Palmer, A. G.; Akke, M.; Ko¨rdel, J.; Rance, M.; Chazin, W. J. J. Magn. Reson., Ser. B 1993, 102, 253-264. (40) Mosteller, F.; Tukey, J. W. Data Analysis and Regression. A Second Course in Statistics; Addison-Wesley: Reading, MA, 1977. (41) Luz, Z.; Meiboom, S. J. Chem. Phys. 1963, 39, 366-370.

2872 J. Am. Chem. Soc., Vol. 122, No. 12, 2000

Millet et al.

in R and R′ were obtained by propagation of the uncertainties in Rex and ∆R2(0.016 ms-1, 1.00 ms-1), respectively.

Results Backbone amide 15N spin relaxation rate constants were determined for 44 residues in BPTI whose resonances were not overlapped in the two-dimensional 1H-15N correlation spectra at any of the temperatures (300 and 313 K) or fields (11.7 and 14.1 T) used. All measured rate constants are provided in Supporting Information. Backbone amide assignments were taken from Glushka et al.42 Relaxation rate constants were recorded using three different pulse sequences, shown in Figure 3, to obtain the widest range of τcp values with optimal sensitivity. The average experimental uncertainty for the experiments of a, b, and c in Figure 3 averaged over all combinations of temperature and static field strengths were 0.10, 0.09, and 0.33 s-1, respectively. A series of control experiments were performed to ensure that relaxation rate constants obtained from the different experimental methods were commensurate. Both a and b in Figure 3 are CPMG experiments; however, different strategies are used for averaging the relaxation contributions from in-phase and antiphase coherences. The accuracy of the experiment of Figure 3b relative to the experiment of Figure 3a was assessed by recording a second relaxation curve at T ) 300 K and B0 ) 11.7 T using the experiment of Figure 3a and τcp ) 10.8 ms. The mean pairwise deviation between the data recorded using a and b in Figure 3 was 0.16 s-1 with a standard deviation of 0.29 s-1. The value of R2(1/τcp) for the smallest value of 1/τcp (τcp ) 64.5 ms) was obtained with the Hahn echo sequence of Figure 3c as opposed to the CPMG sequences of a and b of Figure 3. The reproducibility of the Hahn echo experiment was assessed by calculating paired differences between duplicate spectra. For data recorded at T ) 300 K and B0 ) 14.1 T, the mean difference was 0.17 s-1 with a standard deviation of 0.15 s-1. The standard deviation divided by 21/2 agrees well with the average experimental uncertainty of 0.08 s-1 obtained by the error analysis method described in Materials and Methods. For data recorded at T ) 313 K and B0 ) 11.7 T, the mean difference was 0.06 s-1 with a standard deviation of 0.25 s-1. The standard deviation divided by 21/2 agrees well with the average experimental uncertainty of 0.15 s-1. The accuracy of the Hahn echo experiment was tested by comparing the relaxation rate constants obtained using b and c of Figurs 3. The mean difference, R2(1/τcp ) 0.016 ms-1) - R2(1/τcp ) 0.092 ms-1) was 0.44 ( 0.45 s-1 averaged over all residues not subject to conformational exchange linebroadening (for this calculation, residues 12-18 and 35-42 were excluded as potentially subject to exchange effects). The majority of backbone 15N sites in BPTI are not subject to chemical exchange processes, and the relaxation rate constants R2(1/τcp) were independent of τcp and of the pulse sequences used to measure R2(1/τcp). This result is illustrated in Figure 4 for Gln31. The standard deviation of all R2(1/τcp) values with respect to the mean value for each residue at each temperature and static field strength was 0.28 s-1. This value represents both the within-experiment and experiment-to-experiment variability and is 1-3-fold larger than the random uncertainty associated with individual relaxation rate constants recorded for individual values of τcp given above. Strong dependence of R2(1/τcp) on 1/τcp is observed for residues in the vicinity of the Cys14-Cys38 disulfide bond in (42) Glushka, J.; Lee, M.; Coffin, S.; Cowburn, D. J. Am. Chem. Soc. 1989, 111, 7716-7722.

Figure 4. Relaxation rate constants for Gln31. Values of R2(1/τcp) are shown for (b) T ) 300 K, B0 ) 11.7 T; (O) T ) 300 K, B0 ) 14.1 T; (9) T ) 313 K, B0 ) 11.7 T; and (0) T ) 313 K, B0 ) 14.1 T. Lines are drawn at the average values of the rate constants at 300 K and at 313 K, 7.34 s-1 and 5.51 s-1, respectively and simply serve to guide the eye. Small differences between relaxation rate constants at 11.7 and 14.1 T arise from field dependence of chemical shift anisotropy. The average uncertainty in the relaxation rate constants is 0.12 s-1, and the average standard deviation for the relaxation rate constants with respect to the average values is 0.21 s-1. The former reflects withinexperiment variation and the latter reflects both within-experiment and between-experiment variation.

Figure 5. Relaxation dispersion curves for (a) Ala16 and (b) Gly36 at 300 K. Values of R2(1/τcp) at (b) B0 ) 11.7 T and (O) B0 ) 14.1 T are shown for each residue. The lines are the best fit of eq 19 to the data. Fitted parameters are given in Table 1.

BPTI, including residues Ala16, Gly36, Cys38, Arg39, and Ala40. As discussed previously, residues Cys14 and Lys15 are subject to exchange broadening, but decay of the relaxation dispersion curves was not observed in the present experiments.14 To obtain estimates of Rex, dispersion curves recorded for each combination of temperature and static magnetic field strength were fit independently. In most cases, eq 19, the fast-limit approximation, gave adequate fits to the experimental data, and the full form of eq 3, which requires an additional adjustable parameter, offered no advantages. However, the relaxation dispersion curves for Arg39 at T ) 300 K could not be fit acceptably by eq 19 and the dispersion curves were fit with eq 3. Dispersion curves are shown in Figure 5 for Ala16 and Gly36 at T ) 300 K. Rex values for Ala16 and Gly36 were 1. Calculations were performed assuming Ra ) Rb and pa ) 0.95, but the results depend weakly on these assumptions.

than at B0 ) 11.7 T, and ∆R2(0.016 ms-1,1.00 ms-1) will be underestimated at B0 )14.1 T compared with B0 ) 11.7 T. Improved estimates of R′ are obtained by using R2(1.00 ms-1) measured at B0 ) 11.7 T for calculating ∆R2(0.016 ms-1,1.00 ms-1) at both values of B0. For example, values of R′ calculated by this approach for Cys38 and Arg39 are 0.24 ( 0.23 and -0.10 ( 0.20 at T ) 300 K and 2.10 ( 0.20 and 1.27 ( 0.16 at T ) 313 K, respectively, and agree better with values of R given in Table 3. Relaxation interference experiments also can be used to obtain an estimate of the transverse relaxation rate constant for in-phase coherences that does not contain contributions from chemical exchange effects.46,47 To first approximation, the average relaxation rate constant for in-phase and antiphase coherences is the in-phase relaxation rate constant plus one-half of the amide proton longitudinal relaxation rate constant. Combining relaxation interference experiments with experiments for measuring the amide proton longitudinal relaxation48 may provide an alternative approach for obtaining R2(1/τcp f ∞) for determination of R′. Most investigations of molecular dynamic properties of macromolecules using spin relaxation measure the transverse relaxation rate constant using CPMG experiments with a single value of 1/τcp g 1 ms-1.5 In some of these studies, quadratic field dependence of the exchange contribution to R2(1/τcp) has been observed and taken as evidence for fast-limit chemical exchange kinetics.21,23,24 However, the different τcp-dependence of the dispersion curves predicted by eq 18 for slow and fast chemical exchange complicates the interpretation of such measurements. Figure 12 compares the static magnetic field dependence of Rex given by eq 8 and the apparent exchange broadening obtained by the conventional approach, ∆R2(1 ms-1,∞). In accordance with the usual experimental practice, the figure shows the ratio of the exchange contributions at two magnetic field strengths, 11.7 and 14.1 T. The main result is (46) Tjandra, N.; Szabo, A.; Bax, A. J. Am. Chem. Soc. 1996, 118, 69866991. (47) Kroenke, C. D.; Loria, J. P.; Lee, L. K.; Rance, M.; Palmer, A. G. J. Am. Chem. Soc. 1998, 120, 7905-7915. (48) Peng, J. W.; Wagner, G. J. Magn. Reson. 1992, 98, 308-332.

Millet et al. that apparent quadratic scaling, indicated by a ratio of (14.1/ 11.7)2 ≈ 1.44, is obtained for all values of kex/∆ω provided that ∆ω e 1.5/τcp. In practical terms, the apparent field dependence of R2(1/τcp) measured for a single value of 1/τcp under conditions of rapid pulsing (τcp e 1 ms) does not provide any evidence for the time scale of chemical exchange. As shown by eq 16, at least two measurements of R2(1/τcp) with τcp as long and as short as possible are needed at each field to characterize the chemical shift time scale for an exchange process. The present data comprise an extensive set of relaxation dispersion decays as a function of static field and temperature for BPTI. For the residues with the largest exchange contributions to transverse relaxation, Cys38 and Arg39, dispersion data at both B0 ) 11.7 and 14.1 T were fit globally to provide estimates of the exchange parameters, pa, ∆ω, and kex. The fitted values of ∆ω for Cys 38 and Arg39 given in Table 4 are in excellent agreement with values of ∆ω ) 0.54 × 103 s-1 (86 Hz) and 1.15 × 103 s-1 (183 Hz), respectively, estimated by Wu¨thrich and co-workers from zz-exchange spectra recorded at T ) 289 K and B0 ) 11.7 T.27,49 The values of pa given in Table 4 also agree with pa ) 0.96 estimated by Wu¨thrich and co-workers. In contrast to earlier reports using data recorded at a single static magnetic field,14,28 Cys38 and Arg39 exhibit very similar values of kex ≈ 450 s-1 at 300 K and 1340 s-1 at 313 K when data from two static magnetic fields is analyzed globally. The activation barrier for the reverse exchange reaction, obtained from the temperature dependence of kex using the Arrhenius equation, is ∼64 kJ/mol. These values agree with apparent activation energies for conformational exchange measured previously in BPTI at lower temperatures using zzexchange spectroscopy.27 While the similar exchange rate constants observed for Cys38 and Arg39 support the assumption of a two-state exchange process involving isomerization of the disulfide linkage, kex or ∆ω for Cys14 and Lys15 must be very large, because elevated values of R2(1/τcp) were observed at all τcp values for these residues. This observation suggests that more than one kinetic exchange process may be active in BPTI.29,50 In general, simultaneous fitting of dispersion data obtained at multiple static magnetic fields will be essential for interpreting exchange phenomena outside of either the slow or fast limits on the chemical shift time scale; however, such analyses can be hindered by small magnitudes of Rex, lack of data over a sufficient range of 1/τcp values, and violations of the assumption of a two-state exchange process. Conclusions Chemical exchange is a ubiquitous phenomenon in NMR spectroscopy that is used for characterizing conformational and kinetic dynamics in molecules ranging from small organic molecules to biological macromolecules. In many cases of practical interest, particularly for macromolecules, the site populations of exchanging chemical species are highly unequal as a consequence of the Boltzmann distribution. Under these conditions, the static magnetic field dependence of the chemical exchange contribution to the transverse relaxation rate constant (49) Wider, G.; Neri, D.; Wu¨thrich, K. J. Biomol. NMR 1991, 1, 9398. (50) Beeser, S. A.; Oas, T. G.; Goldenberg, D. P. J. Mol. Biol. 1998, 284, 1581-1596. (51) Shaka, A. J.; Barker, P. B.; Freeman, R. J. Magn. Reson. 1985, 64, 547-552. (52) Kay, L. E.; Keifer, P.; Saarinen, T. J. Am. Chem. Soc. 1992, 114, 10663-10665. (53) Marion, D.; Ikura, M.; Tschudin, R.; Bax, A. J. Magn. Reson. 1989, 85, 393-399.

Definition of NMR Chemical Shift Time Scale varies as δRex/Rex ) R(δB0/B0), in which 0 e R e 2 for pa > 0.7. The value of R depends on the NMR chemical shift time scale for the exchange process. Slow exchange (kex/∆ω < 1) is defined by 0 e R < 1, intermediate exchange (kex/∆ω ) 1) is defined by R ) 1, fast exchange (kex/∆ω > 1) is defined by 1 < R e 2. Therefore, the static magnetic field dependence of Rex determines the chemical shift time scale even if the site populations are so highly skewed that the minor resonance is not observable in the slow exchange limit. The range of values for R observed in basic pancreatic trypsin inhibitor demonstrate that simple assumptions about the field dependence of conformational exchange are frequently erroneous. The scaling factor R is a unique parameter characterizing chemical exchange in NMR spectroscopy and is essential for the interpretation of spin relaxation data acquired at multiple static field strengths. Acknowledgment. We gratefully acknowledge a discussion with D. A. Torchia (NIH) at the 40th ENC that stimulated our interest in the problem of chemical exchange in systems with skewed site populations. We also acknowledge discussions with

J. Am. Chem. Soc., Vol. 122, No. 12, 2000 2877 M. Rance (University of Cincinnati) that clarified the theoretical results and with L. E. Kay (University of Toronto), concerning the global analysis of dispersion curves. We thank D. A. Torchia and R. Ishima for a preprint of reference 34. O.M. was supported by a predoctoral fellowship awarded by the Ministerio de Educacion y Ciencia (Spain), J.P.L was supported by a National Institutes of Health postdoctoral NRSA (1F32 GM 19247), and C.D.K. was supported by a National Institutes of Health Training Grant (T32 GM08281). M.P. acknowledges a grant from the Direccion General de Ensen˜anza Superior (PB 97-0933), and A.G.P. acknowledges a grant from the National Institutes of Health (GM 59273). Supporting Information Available: Four tables containing the measured values of R2(1/τcp) at static magnetic field strengths of 11.7 and 14.1 T and temperatures of 300 and 313 K for basic pancreatic trypsin inhibitor (PDF). This material is available free of charge via the Internet at http://pubs.acs.org. JA993511Y